Research On Speaker Recognition In Noisy Environment

Speaker recognition deals with recognizing the identity of the person speaking utrerance,It is the process of automatically recognizing who is speaking based on the information obtained from the speech.Speaker recognition has a wide range of applications which include banking or credit card transctions by telephone,information and reservation services,access control in high security areas and forensic investigations.Though speaker recognition systems perform well when clean speech is used for training and testing,the performance degrades rapidly when speech used in real-world conditions.The focus of this research effort is to develop techniques for noise cancellation with emphasis on the problem of speaker recognition in noise which include speech endpoint detection,speech enhancement,feature extraction and back-end Processing,and some research results were obtained.A history review of speaker recognition theory is introduced in this paper at first. Based on the result come from the research with the speaker recognition,hotspot and nodus of the research are list.For speech signal has the characteristics of chaos, fractal dimension theory is a kind of means to describe chaos signal. According to simple computation, good anti-noise ability and low precision of Katz algorithm and complex computation and good precision of box-counting dimension, a improvement fractal algorithm based on wave (IBW) was presented and analyzed with the fractal Brown curve and noisy speech compare with box dimension and Katz dimension .The theory analyse and experiment showed that IBW-FD has lower computation and higher precision than Katz dimension and box-counting dimension.IBW-FD also had stronger ability of anti-noise and distinguish gauss noise with speech than the others.The result showed that IBW-FD was the good speech fractal algorithm because of low complexity, good precision and nice anti-noise ability.As a hotspot of research on speech signal processing,speech endpoint detection is the first steep of speaker recognition systems.According to the characteristics of speech and noise,improvement fractal algorithm based on wave (IBW) is introduce in speech endpoint detection and a kind of new method is proposed which named as VAD way of unite energy and fractal dimension (UEFD).Finding new speech feature and combination of old speech features is hotspot of research on speech feature extraction.This page combine the IBW-FD and MFCC as mix speech feature of speaker recognition.Experiments were conducted that the new mix speech feature shows better performance than former methods.Two new method for Discrete Fractional cosine Transform are proposed and the two method is introduced in speech enhancement.According to the characteristic of speech enhancement,four constructions are proposed as speech enhancement based on two cycles Discrete Fractional cosine Transform(FDCT2),speech enhancement based on modified two cycles Discrete Fractional cosine Transform(MFDCT2),speech enhancement based on three cycles Discrete Fractional cosine Transform(FDCT3) and speech enhancement based on modified three cycles Discrete Fractional cosine Transform(MFDCT3).Compared with DCT algorithm, our algorithms perform...